Regularized total least squares based on quadratic eigenvalue problem solvers (Q2566641)

The authors propose a new iterative algorithm RTLSQEP for the regularized total squares problem via quadratic eigenvalue problems solved at each iteration. The performance of the above method is tested by a Matlab implementation and compared with other regularization solvers. Very few iterations are needed and the proposed method is robust regarding the initial vector of the iterative algorithm. The crucial question is concerned with efficient solvers of the quadratic eigenvalue problem treated here by means of some citations as \textit{R. C. Li, Q. Ye} [SIAM J. Matrix Anal. Appl. 25(2), 405--428 (2003; Zbl 1050.65038)] and given in the paper by \textit{F. Tisseur} and \textit{K. Meerbergen} [SIAM Rev. 43(2), 235--286 (2001; Zbl 0985.65028)] and I am afraid that some numerical troubles should re-appear.